NavList:

"Both I and the boat's captain (who also was doing some sextant work) felt like we noticed two occasions when the GPS clock jumped ahead by a couple of seconds on a given day, and then fell back to what we expected the following day."

My word you chaps seem to work to stringent limits.  I’m afraid I always buy the cheapest watch from the cheapest stall in the market knowing full well I’ll have left it lying around somewhere within a couple of months, but is two seconds going to make that much difference to your position fixing?  One of the problems with statements on Nav List is that people don’t give sufficient information for the statements to be analysed properly by others.  Was this day or night?  Was it a fix or a position line? I’m assuming it was a noon fix.  What was GPS being used for?  Was it to provide a source of time or to provide a position to compare the astro to?  Assuming the time of maximum altitude of the Sun was two seconds out, is that serious?  The Sun travels around the Earth at 15 per hour so two seconds of timing error equates to 0.5 minutes of longitude or 0.5nm at the equator.  I’d have thought that was pretty good and not worth questioning GPS over.  Of course, you might have using "a couple" more loosely than it's exact meaning.

 Turning to GPS, timing in GPS has to be the complete opposite of the above.  As I understand it, synchronisation has to be to the nanosecond for it to work accurately, and this is achieved within your receiver by the inexpensive clock in your receiver being pulled into sync with the expensive clocks in the satellites through the four satellite solution.  GPS runs on GPS time, which is atom based, and which differs from UTC (atom based plus leap seconds) by an integer number of seconds.  UTC shouldn’t differ from Mean Solar Time by more than one second.  Your receiver uses its “firmware” to add those seconds to its GPS time to give you UTC or LMT readout on the front of your receiver. 

 The only way the UTC readout on the front of your receiver could be out by two seconds without mass resignations and sackings at GPSHQ is if you’d only just switched it on and it was out of sync, if there was a problem with the stored offset between GPS time and UTC, or a problem with your “firmware” all of which are extremely unlikely.  DaveP